Anthropic's Labs Lead On Fable's Capabilities + Building AI-Native Products — With Mike Krieger

0:00Anthropic is not just a massive model builder, it's a massive product builder as well with products like Cloud Code and Cowork that have taken off like crazy over the past few months. And Cloud Code came out of a place that few of us know much about called Anthropic Labs. And Anthropic Labs is a organization within Anthropic that is working on building the next level of frontier products with AI at the center. And so today we are lucky to hear from the person running that lab, Mike Krieger, who is the co-founder of Instagram and now the lead of Anthropic Labs at Anthropic. We're going to welcome him on stage along with Lauren Good of Wired, who will join me as a co-interview.

0:44Mike and Lauren, let's hear it for both of you guys.

0:49Thank you. You're here. In the face of ongoing disruption and opportunity, TMT leaders need to deliver tangible results, not just ideas. When pace and performance matter most, PwC combines market insights and deep sector experience with AI, cloud and emerging tech to accelerate your transformation and drive measurable ROI from strategy to execution. PwC can help you anticipate what's next, outpace disruption and compete. For more information, visit pwc.com. All right. All right, Mike. So chill times in an anthropic land. Nothing going on. Slow week. Slow week. You want to start, Lauren? Yeah. First of all, I mean, we want to talk about what you're working on at labs and explain your role to folks.

1:41But I want to ask you first, how close are you right now to the situation with the White House? Less than in my CPO role. So I transitioned about like five months ago into this labs role. I think in the CPR role, I think would have been deep in it. Now I'm more, you know, obviously, you know, we want to restore it. And as a like product person, I want to make sure that that gets access, but less close to it than, you know, in that sort of C-level role that I had before. Okay. Alex, do you have a follow-up? Oh, yeah. I have like eight follow-ups. Definitely. Well, there was a guy named Ben on X who said, I will mail Anthropic an original copy of my long-form birth certificate if they will enable Fable for me again.

2:18I sound like those lunatics who are obsessed with 4.0 now. Will you take Ben's long form? I don't know that we'll take his long form, but it has been interesting. I mean, Fable was only available for a few days, but I've definitely, every time I've tweeted since then, they have not read whatever I was tweeting, and they've mostly been like, bring back Fable, which, like, on Instagram, we got rid of Gotham. Do you remember Gotham, the filter? This is like, and then for the rest of, like, the next eight years, all I heard was bring back Gotham, so it struck a nerve. Fable will come back before Gotham did.

2:48But yeah, it's clearly the folks that have gotten to use it and started incorporating it. It's actually really interesting. I've learned to not really trust day of or even week of model reactions. You don't really know until you've put it through its paces. And so I almost just completely block out the noise in the first couple of days of any new model release. Because, I don't know, everybody has maybe their toy example thing that they like to do with a new model. but it's hard to actually put it through its paces until you've actually had real work done with it. And I think people were just starting to do that, and then we had to sort of pull back Fable.

3:22But I remember in December when we put out Opus 4-6, it was like this interesting time where everybody went home for the holidays, and a lot of people had that week off between Christmas and New Year's. And then they came back and were like, oh, I spent a lot of time, and I really get why Opus is good, and I'm going to do it. So I don't think Fable has had that opportunity yet. But despite that, though, I mean, this was a pretty big reaction from the Trump administration. And I think everyone here, especially if you were listening to the Alex Stamos session, understands what's going on. But this happened within a few days of the model release.

3:54If I can give Wired some credit, Wired just reported last night that it was due to Anthropik's relationship with or having given access to the model to SK Telecom that could have raised flags within the administration. how surprised were you by how immediate that that backlash was yeah I think the the the sort of reaction decision was surprising and when we were sort of immediately engaged with them to you know restore access as well um and so at the same time you know the one thing that we we think a lot about internally is you know there used to be a poster on the Facebook wall when we were so that was like every day feels like a week and I think that's becoming true in AI and I think a good thing to remind ourselves of in general in the industry is we're dealing with unprecedented times.

4:38We're dealing with new situations and they can develop really quickly as well. And so I think also developing the capabilities and the connections and to make sure those conversations can happen quickly is really, really important. I have a new motto suggestion for you for the wall. Move fast and jailbreak things. I don't think they're going to use that, Lauren. OK. You know, we just heard from Alex that some of these capabilities have been available on, you know, previous models or, you know, you can you can find bugs with other with other models that are available today. So why do you think Anthropic got singled out on this front?

5:14I don't know why like Fable specifically was singled out as, you know, again, Fable being the non-cyber intention model as well. I think that the thing that does change over time is there's, you know, the capabilities, if you knew what to prompt and knew what you were looking for, and then there's the capabilities, I think, you know, I'm not sure I resonate with the, like, juicing high school athletes metaphor from Alex, but like, you know, the uplift that you get, like, uplift is a thing that we think about a lot when we think about model safety. So if you look at our model cards, for example, one of the ways that we look at risk in the bio domain is comparing the uplift from sort of a lay person using the model versus, you know, an expert or a lay person just using the internet and seeing what the comparison there is.

5:58And so that is one trend that has been progressing as the models get more capable. And so maybe less why Fable is singled out and maybe more like what the overall trajectory is. Interesting. So we just had R. Karazian, the lead economist on Ramp on the podcast. And one of the interesting things that he said was, you know, we had during the Pentagon situation, there were these headlines, okay, this company will not use Anthropic anymore. But actually the data from Ramp shows that spending actually increased to Anthropic models. It was apparently a good publicity moment for the company. So I think it sort of plays into a debate that we have like here on the show about like how much of this is like real concern for the issues and how much of it is, you know, from Anthropic is marketing.

6:42And like we have somebody from Anthropic here who can actually shed some light on that. We just said Stamos talking a little bit about it from his perspective on the security side, but we're lucky to have you here today. So so is it material or is it marketing or some combination? I mean, I think it is one of the hardest things to like really deeply believe that something is true and not marketing and then due to not even just anthropic. I think like, you know, I think people are generally right to be skeptical of any company saying anything and you should like put it through your own filters as well.

7:14But for me personally, I was like, no, but it's real. And like we like both deeply care about safety and are like To the extent that we are being vocal about anything, it is to either sort of help paint the picture of what very likely is coming or we believe is coming or what we've already seen and spotted. For example, in the Mythos case, really just looking at vulnerability scanning and bug finding and doing it in partnership with companies that were in that kind of project glasswing initial announcement. And so the technology is awesome. awesome, not in the, like in the just, you know, it is doing really incredible things.

7:50And therefore, even calling out what we see as what is happening, I think can seem hypey. I wish I could press a button and I could make everybody believe that we are not being hypey. I realize that's not the reality that we operate in, but at least from my perspective, we try to call it like it is. My understanding is that within labs in particular, and research and development at Anthropic, that you're using the best AI models to actually build new products, to prototype new products, and sort of test out your thesis. So is this ban essentially now limiting your ability to do that within labs?

8:25Yeah, I mean, definitely Fable is the best model I've ever used. And it's not to say that work has stopped, but it's definitely less good than the other models that are using models that are less good than that as well. And it is also, I mean, maybe the reverse of the distrusting the first week, you know, response is what happens when you don't have Fable. And like, obviously the Twitter reaction for the people that had already kind of like gotten into the model and were using it, right, was strong. But I'd say even in my personal use, I'm like, oh, like I'm on Opus 4.8 and it's good. Like I'm still productive, I'm doing work.

8:59But, and we can go into sort of like how my work changed with sort of these like Fable or the models of that sort of family, but it is noticeable for sure. Yeah, I think we'd like to know that. I mean, the public had access to Fable for like a half a minute. Groups in this like Project Glasswing have had access to Mythos. We don't really know what the difference is between using a model that you can use today and using one of these super anthropic models. So what actually could you do differently with a Fable or a Mythos? I think for me, it's the sort of scope and scale of delegation. And all these things are really imperfect.

9:40Like, you know, people say like, oh, is this now a level five software engineer or level six? But anybody who's used these models extensively knows that they're still spiky in capabilities, right? In some ways, they're, you know, in many ways, they're a better engineer than me. And in other ways, like, you know, I was complaining today that it had missed a descender, like the G, that bottom part is called the descender. I'm like, how did you put that in the UI? And it's clipped in And of course, like there's vision capabilities that need to improve and there's debugging capabilities. And there's sometimes even just sort of human common sense that is, you know, we're way better at than the models are.

10:13But overall, I think the big shift for me working, and it was really interesting because it sort of coincided with me going back into a builder role. So I really got to see going from, you know, using these models as an executive, you're trying to do the most of them, but, you know, not going to have it right. all your email and I think strategy still needs to come from you and then you can use the models to sort of pressure test it. But going back into a builder role and going from, okay, I am delegating chunks like, please fix this bug or I'm thinking of implementing this feature, let's go back and forth to something that ends up being much more sort of, all right, I got this bug report from one of our users or I have this notion of something that I want to build, like, can you sketch out two or three ways in which we could do it?

10:57right like that seems plausible often I find actually sometimes it'll uh give me the the sort of explanation or proposal and be like okay that actually is over my head like you are clearly way smarter than me like explain it to me like I'm maybe not five but at least you know uh uh not you and it'll sometimes you know uh explain it that way but then go build it and get it right getting it right like you know at a very very very high rate and I think that starts really changing how you operate like I moved much more to before going to bed making sure that I had like queued up for Fable, like enough chunky work to last, I would call it the whole night and I'll check in later and it got it done in an hour.

11:33And it was like, they're just hanging out for the next seven hours, but like really like delegating like a, a much more of a goal than just a, so one, one task, for example, like give us an example of one task you would hand to it. Uh, I mean, here's a kind of crazy one, which is, uh, for the programmers in the audience, like I, uh, I had written one of our labs, uh, projects in, um, in Python. That's like the language I You know, all Instagram was all Python. And for like some not super exciting reasons, we actually needed it to be in TypeScript to deploy it. And I was like, all right, that's going to be like, you know, in Instagram, we for years talked about moving from Python to PHP or hack or the Facebook language after the acquisition.

12:10And at least when I was there, never did. But I basically, we have a feature called dynamic workflows where you can have it like also break down the task into like a lot of subtasks. And I trusted it to sort of not just do the individual action, but here's like a whole language conversion of millions, hundreds of thousands of lines of code at that point. Go off and do it. Go plan it. Go execute it. Go verify the work. Double verify the work. And then I came back to the work being complete. So that level of like, this is a big sort of chunky task. So that was, so you're basically saying it was faster.

12:41Did it in an hour? You're guessing compared to, with Fable, compared to what it would have been before. I think the main difference is in the past, it would be like, great, I did it. and you'd be like, well, did you, you kind of took a shortcut here, or this is not quite right, or I need to go verify it, or like, oh, you cut this corner, or. It's like the managing interns thing that everyone's been saying for the past year, which is very offensive to interns, by the way, but yes. Yeah, exactly. I don't know, have you managed an intern? Yeah, that's true. And it was, and you're saying it was more correct.

13:09It was, so it's faster, it's more accurate, more reliable, and then according to the U.S. administration, dangerous. I think the other piece is it has like a greater, theory of mind is the wrong word, but sort of like theory of project so that it's less, you know, oh, I'm going to make this change and it'll say, great, I'll make this change. But really, especially if you've done like software engineering at scale, the best engineers kind of keep in mind all the disparate parts of how this thing's, and they also see around the corners, like I can make this change, but if I don't do it in this way, then the next change is going to be incrementally harder.

13:41And I think that's like been a significant difference I've seen in that kind of class of models. So I think when we talk about Anthropic Labs, right, people think of Claude Code because it is really your breakout product. And it sounds like you've been tasked with basically figuring out what the next Claude Code is. Would you say that's an accurate description of what you're doing at Labs? And also, why does Anthropic need Labs? Labs, yeah. It's also maybe worth thinking about why we needed Labs in 2024 when I arrived and why we need labs today because I think that the answer kind of shifts. I started the original labs team with Ben Mann, who's one of the co-founders of Anthropic, in my third week at Anthropic, and it had been something that I'd been bubbling under.

14:23And at the time, the reason was really different. It was all of our product engineering team was 25 people, and we didn't have the models, really. We had Claude, when I joined, was Sonnet 3, like you, and Opus 3. Those were, for the time, good models, but you weren't going to, they were not even interns, right? They weren't even IC3 engineers. So if you have a team of only 25 or 30 engineers, they are working on like the next incremental thing. And we were feeling like the models are starting to get better, but we don't have any products that sort of show that off. Like a good litmus test for me is when we get ready to release a model, do we have either a product or a demo or some other illustration of something that is very different?

15:02And it gets harder over time. Like with Fable, you know, even like illustrating like that weekend task or, you know, this longer amount of work. So really, labs at the time was, let's make sure we don't, like, our products don't fall behind the model exponential that's happening. And yeah, so Cloud Code came out of that initial one because nobody in the rest of the product orgs, people were thinking about coding, but nobody was sort of had the, like, space to go and think about, well, what if we totally change the form factor and we embrace the fact that the models were going to evolve in this way?

15:32And a lot of the two most useful thought exercises we do in labs, one is visualize the gap between what the models can do today and how most people use it. It can be closed that gap. That's one. And the other one is imagine what the models are bad at now that they're actually going to be really good at in six months. And let's make sure we have a product ready for that by then. I think those are the two guiding questions for labs. And then also out of that first incarnation came computer use. Computer use was different, though, because when we built it, it was really bad. Like we tried a bunch of products with it and this was around, you know, Sonic 3.5 and be like, Claude, can you help me, you know, clean up my desktop?

16:07And it would like click the thing, delete the file. Like this is not safe for release. We're definitely not going to go and build this or to ship this. But we had that product so that every new model that we'd release, we'd first check it internally and say, the computer used to get better. And we'd tell the research team how it had gotten better or worse until the moment where we said it's good enough. we're actually going to put a product out around this. It also gives you this sort of beacon into the future that you can measure your future products against. But then compare it to now. So we have a thriving product team.

16:38There's Cowork. There's, you know, Claude Code has grown a lot. We have our platform. And now I think it's actually much less about none of these product teams are doing this sort of thinking. And I think it's much more that the models are advancing really quickly. and even our capability to interact with them needs to evolve. So one of the things we collaborated with labs and CloudCode that we shipped today is CloudCode Artifacts. Having CloudCode not just be able to type back to you, but also sort of draw a picture or give you an illustration. And that partially came from spending a lot of time in labs saying, just a text box and a big text response is not going to cut it anymore.

17:14Like when I mentioned that the models feel like they're way smarter than me when they talk to me, sometimes I'm like, can you draw me a picture? Because this is what I actually need to fully understand this. But it's really what we've been thinking about is, you know, yes, we have a lot more products. You know, we actually have a lot of consolidation to do in our products. Like, that's another initiative that we have. But within that, we still have an opportunity to make things much more accessible to a person that does not spend all of their time thinking about prompting and the exponential and the difference between high, low, and medium effort.

17:42Like, there's a lot we can still do there. Right. But, Mike, so there's a – it puts people using anthropic models in an interesting place, right? Right. You know, Cursor, I think, just sold for 60 billion to SpaceX. And someone put this meme on Twitter that, like, you know, Cursor would have sold for 300 billion if it wasn't for this guy. And it's a picture of Boris Churnity, the person who created Claude Code. And so for companies that are going to build on top of Anthropik technology, you know, they're going to wonder, do I want to partner with Anthropik or is Anthropik going to go ahead and build the product that I'm going to want to build, potentially even after partnering with them.

18:20Yeah. I mean, we'll take the agentic coding side, and I think the broader sort of aspect of being both a platform and a product I think is really interesting. When we take on projects, the goal is often to sort of push that area of the industry forward. So there were AI coding editors, and some of them were really good, but nobody was quite thinking about it in as sort of freeform a way as we got to think about it with Cloud Code. And now a lot more products have that flavor than I think would have otherwise. And so I think if we're ever, you can call me out on this, Alex. If we're ever entering an industry where you're like, all you're doing is the same thing everybody else is doing, but you've got the Anthropic brand.

19:00I feel like that's a bad use of our time and a bad use of our either labs or product team time. If we're going in somewhere, it should hopefully be to say, all right, we think that the direction of travel is this way. We can build a product of that. And then, by the way, there's no world, nor should there be a world where all the products are Anthropic products. That would be a bad world, right? So that is hopefully either creating new space for companies or sort of showing the way where other products can incorporate. Yeah. It would almost be like working for a tech company that has social messaging, video, if all.

19:30Right, Mike? Yeah. Okay. You did leave. Well, there was some question, for example, when Anthropic launched a product that was seen as competitive to Figma and you had been on the Figma board prior to that, and I think you stepped down. I did, yeah. Is that correct? Yeah, and so it's a good question that Alex has brought up. I think where Silicon Valley is known for this really healthy, vibrant, risk tolerant startup ecosystem. And when the big start coming in with tons of venture capital and, you know, a lot of resources, people say, well, wait, are they just are they essentially just going to steal my idea?

20:01Yeah, no, I think I think our dual existence and it's something that other companies have to navigate. We talked about Amazon a lot in the in the previous panel, like they have to navigate this role where they are both the infrastructure provider. They obviously have a very large e-commerce thing. They do video, but they also serve video. And then, you know, by and large, customers can live in that dual world of like, okay, I'm using their infrastructure, also knowing that they are also using their infrastructure to do that. And I think the, you can talk to our customers and see how well we're doing at this.

20:30Like the thing I always try to do is like at least approach it with a lot of transparency. So the cursor example is an interesting one where like Michael and I talked a lot over the, you know, time around here's where things are heading. And, you know, similarly with the other products that we think about, like, can we, I think it's a couple of things. It's transparency and then it's shared building blocks. Like, yeah, I think in general, and I actually don't think there's any cases where this is even true. Like we're trying to build on top of the same capabilities that are available elsewhere.

20:57The last time I was here in the Commonwealth Club on this stage was our healthcare day at the beginning of the year. And we didn't ship like Claude healthcare, only we have it like nobody else has it. We shipped a bunch of like plugins and skills and MCPs and like complimentary abilities. So that's how, I'm not claiming it's easy or that it's a straightforward thing, but it is how we're trying to navigate what is like admittedly a complicated sort of situation. Speaking of startups, Anthropic is still technically a startup, but you're worth a lot of money. I mean, what's the latest valuation? Is it?

21:28965. 965 billion dollars or something like that. You sold Instagram for a billion, right? Startup. In 2010. Right. Financials have changed quite a bit since then. And yet, Anthropic has positioned itself. It is a PBC and it's positioned itself as sort of a more ethical company around building AI. And I'm wondering if you could talk a little bit about how you see that positioning and Anthropics role in particular, changing the culture of the valley. Like, I know I think back to how Google in the beginning of the 2000s really changed the culture of Silicon Valley in so many ways. And how do you see Anthropics culture now dictating this next era?

22:08Yeah, that's a really interesting question. Maybe I'll start like insight. And I think there's an external component to I think the reason I joined in the first place, so I was winding down my second startup, and knew I wanted to go work at a frontier lab, because I'd started to use these models for coding. And they were bad at coding, but I could see that they were as bad as they were ever going to be at coding, they were going to improve. And I had started building on top of these API's. So the startup I was doing was called Artifact. And we did sort of AI powered sort of news recommendations.

22:35And she read a lot of big technology via Artifact back in the day. was one of the things we added. But not wired. You know, you guys had a really hard paywall, to be honest. Fair enough. We didn't do great on paywalls. Do you need a discount on my subscription because I can get one for you? Okay. It's actually really funny. Making deals. Yeah, making deals. It's the login cookies. It's really hard to keep people. I know, I know. Please escalate this to Condé now. I know. And email login's very hard to do in an app. But I was building on top of the APIs and be like, wow, okay, they're able to do really interesting things.